Technical approach

The measurement problem

Vision infers contact from geometry and deformation cues at a distance. Tactile measures directly at the contact interface. These are different epistemic categories.

A camera sees a hand closing around an object. It estimates contact from visual features: the grip aperture, surface deformation, object pose. The inference holds while the contact surface is visible. As soon as the hand occludes the object, those inferences degrade.

Tactile sensing measures force distribution at the contact surface directly. The signal does not depend on visibility. It is a different measurement modality from vision, and it remains stable under occlusion.

The four-stage stack

OVR builds a complete path from physical contact to control loop input.

• Sensor array. A piezoresistive composite layer converts normal and shear force at each taxel into resistance changes. Spatial resolution depends on electrode pitch and taxel count.
• Readout circuit. Embedded electronics sample the resistance matrix at frequencies useful for real-time control, delivering a digitized force stream.
• Data pipeline. The raw signal stream is packaged for consumption by motion controllers and planning systems. Format and transport are designed for integration with existing robot software stacks.
• Control loop integration. Tactile signals enter the control loop as measurement inputs. The planning system receives additional information about contact state. The motion controller receives feedback for force-regulated tasks.

Scope boundary

OVR builds the sensing and control layer.

What OVR does not build: the robot arm, the vision system, the motion planner, or the reasoning model. The tactile layer is designed to integrate with existing hardware and software stacks through a hardware-agnostic interface path.

If you already have a capable manipulation system that fails on contact-critical tasks under occlusion, OVR's layer adds measurement where vision cannot see.